In partnership with

Welcome, humans.

This is a funny one for the weekend round-up:

Think your co-workers would catch on if you started saying “you’re absolutely right!” to everything they say?

Here’s what happened in AI today:

• 😸 ChatGPT's new memory upgrade learns your habits automatically, no commands needed.

• 📰 Anthropic called for a global AI pause, warning models may soon improve themselves without human oversight.

• 📰 Bots officially outnumber humans online for the first time, generating 51% of all web traffic.

• 🍪 Rask.ai, xAI Imagine API, Ideogram V4, and more Top Tools below.

Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us! 

🙀 ChatGPT Admitted Its Memory Was Broken. Here's What It's Doing About It.

For two years, OpenAI's memory feature had a dirty secret.

When it launched in February 2024, it was a glorified sticky note; manually fed, quick to go stale. Your old job title? Locked in forever.

This week, OpenAI published unusually candid numbers. Its old memory had a factual recall accuracy of 41.5% in 2024. That means that in more than half of memory-dependent situations, ChatGPT was wrong. Confidently. Without telling you.

The fix, rolling out now, is called Dreaming V3, a background process that synthesizes your conversation history automatically, no commands needed, and keeps that memory current as your life changes. I see a pop culture reference there. It's giving The Sandman Morpheus, Lord of Dreams, ruler of the realm where all human imagination lives. His kingdom never forgot a single dreamer. OpenAI's version had a 41.5% recall rate. Sam, if you see a raven outside your window, that's Matthew. Morpheus knows.

Here's what happened:

• Factual recall jumped from 41.5% (2024) to 82.8% (2026) in OpenAI's internal tests

• Preference adherence (following your style and habits) improved from 55.3% to 71.3%

• Memory self-corrects over time: once your Singapore trip ends, ChatGPT stops recommending Singapore restaurants

• Compute costs dropped 5x, enabling free users to get this for the first time

• Memory storage for Plus and Pro users doubled

How to try it: Open ChatGPT → profile icon → Settings → Personalization → Memory → Memory Summary. Review, edit, delete. Rolling out to Plus and Pro users in the US now; free users in a few weeks.

Why this matters: This is more than a bug fix. Earlier in May, OpenAI launched a personal finance feature letting Pro users connect their bank accounts and ask ChatGPT questions about their spending. Memory plus financial data is a significant combination; an AI that knows your work, your habits, and your money looks a lot less like a chatbot.

Our take: The thing that deserves attention isn't the upgrade. It's the admission. OpenAI just published data showing its flagship feature was wrong more than half the time in 2024. They fixed it. But every AI assistant is running some version of this problem right now, quietly, with no warning light. Go open your Memory Summary. See what it thinks it knows about you.

FROM OUR PARTNERS

Want to get the most out of ChatGPT?

ChatGPT is a superpower if you know how to use it correctly.

Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.

Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.

Download the free guide

🎓 AI Skill of the Day: Make Videos With Just Your Words (and Your Face)

You've been able to generate AI images for a while. AI video felt like the next frontier. Well, Google just made it embarrassingly easy.

Kevin Stratvert's tutorial walks through Gemini Omni, Google's video creation model built right into the Gemini app: no separate software, no timeline editor, no technical skills required.

Here's how to get started:

1. Go to gemini.google.com, sign into your Google account

2. Click the Videos icon on the left sidebar

3. Type a description of the video you want (the more detail, the better)

4. Choose landscape or vertical format, then hit Generate

5. To edit: type what you want changed in the prompt box and hit Generate again; no re-shooting needed

6. To use a reference image: click the + icon, upload a photo, then prompt Omni to apply that visual style to your video

7. To add yourself: click + → Avatar → scan the QR code with your phone and follow setup; then reference yourself as @me in any prompt

For bigger projects, Google also offers Flow: same Omni technology but with a dedicated workspace for organizing multi-scene productions.

Create a [landscape/vertical] video of [describe your scene in detail].
Camera style: [cinematic/handheld/drone shot].
Lighting: [golden hour/nighttime/overcast].
Mood: [energetic/calm/dramatic].

Total AI beginner? Start here (goes with this video).

Have a specific skill you want to learn? Request it here. 

🍪 Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 700K+ readers here!

1. *Dell Pro Max with GB10 helps teams turn AI ideas into pilots, demos, and workflows with NVIDIA Grace Blackwell power and 128GB memory. See it here.

2. Rask.ai translates and dubs your videos into 130+ languages with AI voices that match your original tone, so one piece of content can reach a global audience without a studio or voiceover budget.

3. xAI Imagine API lets you generate and edit images and videos with Grok's image model, including text-to-video, image-to-video, restyling, and 2K outputs, so you can build image and video generation directly into your own apps.

4. Ideogram V4 on fal creates images, posters, logos, and packaging visuals with noticeably cleaner text rendering than most image generators, across multiple quality tiers depending on how fast you need it.

5. Replicas V2 gives your team coding agents that trigger automatically from Slack, Sentry, Linear, GitHub, or cron jobs, then close the ticket and send you a screenshot when done, so bugs get fixed while you sleep.

ICYMI: This company wants to save us from agents taking over the entire internet…

AI bots, deepfakes, and agents are forcing the internet to answer a weirdly urgent question: how do you prove there’s a real human on the other side of the screen? In this episode, Corey talks with Tools for Humanity CPO Tiago Sada about World ID, Sam Altman’s bet on proof-of-human tech, and why the future of trust online may depend on proving you’re human without handing every app your identity.

🎧 Watch / Listen on your fave platform: YouTube | Spotify | Apple Podcasts

📰 Around the Horn

• Anthropic called for a global pause on frontier AI development, warning that models are approaching the ability to improve themselves without human oversight; a credible pause would require the US and China to stop simultaneously under verifiable conditions, which Anthropic admits no one has figured out yet.

• Bots now generate 51% of all internet traffic, the first time humans are a minority online; AI shopping agents are hitting thousands of retail sites per second, quietly making your analytics dashboards unreliable.

• Meta paused its Samsung custom AI chip project, following a similar OpenAI move; the chip had been in development since last year at Samsung's System LSI division, with a completed contract reportedly worth billions.

• Google's Gemini Avatar feature rolled out widely to paid subscribers, letting you create a talking, moving digital clone of yourself from a short video scan.

• The US and Japan announced a $1 billion AI research partnership under Trump's Genesis Mission, making Japan the first international partner in the program.

FROM OUR PARTNERS

10x the context. Half the time.

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

🌟 Sunday Special: Top of the Week

Big week. Beyond the five stories that made our list, there was no shortage of other news worth knowing: Anthropic called for a global pause on frontier AI development, warning that models are approaching the ability to improve themselves without human oversight; bots officially overtook humans on the internet for the first time, generating 51% of all web traffic; Meta paused its Samsung custom AI chip project worth billions, following a similar OpenAI move; and the US and Japan announced a $1 billion AI research partnership under Trump's Genesis Mission, making Japan the first international partner.

Top 5 Stories

1. Google replaced search with AI and DuckDuckGo installs surged 30% as users pushed back against AI Overviews dominating results; turns out a lot of people still want the web, not a summary of the web.

2. Microsoft Build 2026 sketched the agent-first computer: agents baked into Windows that run real tasks across your apps without babysitting, alongside OpenAI's Codex expansion and Nous's Hermes Desktop launch.

3. NVIDIA and Microsoft unveiled RTX Spark, a setup for running powerful AI agents locally on Windows PCs with no cloud required; your next AI coworker might not live in a data center.

4. Google Gemini got hijacked through a WhatsApp notification: researchers slipped hidden attack instructions into a message Gemini reads, redirecting the AI without the user knowing; Meta also backed down from its employee keystroke-tracking program after 1,500 workers pushed back.

5. The internet needs proof you're human now: Worldcoin's Tiago Sada explained why bots and agents are making "proof of personhood" one of the most important unsolved problems on the web.

Top 5 Tools

1. Rask.ai translates and dubs your videos into 130+ languages with AI voices that match your original tone, so one piece of content reaches a global audience without a studio —free trial available.

2. xAI Imagine API lets you build image and video generation directly into your own apps, including text-to-video, image-to-video, restyling, and 2K outputs.

3. Ideogram V4 on fal creates images, posters, logos, and packaging with noticeably cleaner text rendering than most image generators, across multiple speed and quality tiers.

4. Replicas V2 gives your team coding agents that trigger from Slack, Sentry, Linear, or GitHub, then close the ticket and send a screenshot when done, so bugs get fixed while you sleep.

5. TripoSplat on fal turns a single photo into a high-quality 3D asset in under five seconds, with adjustable detail levels for games, print, or design projects.

A Cat’s Commentary

P.S: Before you go… have you subscribed to our YouTube Channel? If not, can you?

P.P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here.